The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Voi. XXXVII. Part B7. Beijing 2008
454
(http://spl.bnu.edu.cn): the data of winter wheat is measured in
Beijing, China, in 2001, 2003 and 2004; summer maize is
measured in Hebei province, in 2000 and 2003; cotton is
measured in Xinjiang province, in 2003 and 2004; rice is
measured in Zhejiang province, in 2002 and 2003; and rape is
measured in Zhejiang, in 2003 and 2004.
2.2 “Beijing-1” imagery
The study area stands in Beijing and Tianjin Cites located in
North China Plane under typical continental monsoon climate
characterized by hot and rainy summers, cold and dry winters.
The three scenes of Beijing-1 images (UTM projection, pixel
resolution of 32m) were obtained in 16 March, 2006, 14 May,
2006 and 17 June, 2006, respectively. Radiation corrections
were performed using the radiation correction coefficients
supplied by Dr. Chen Zhengchao in Institute of Remote Sensing
Applications Chinese Academy of Sciences. The major crop
species in the area are winter wheat and summer maize.
The ground-based optical LAI measurements were obtained in
2004 using the instrument LAI2000 from spectrum database of
typical land surface objects. The work in this paper was to
extract the datasets from the Spectral Database System of
Typical Objects in China (http://spl.bnu.edu.cn), analyze
systematically and apply the prior knowledge to remote sensing
scale.
3. METHODS
3.1 The statistical analysis of spectrum datasets
The overall statistical analysis of spectrum datasets depends on
two steps that influence the expression of the prior knowledge.
The first step is to specify the growth stages of crops for every
measured spectral data, because the work of wrongly divided
growth stages will lead to different sample quantity and
disharmony with others’ work. The second step is to calculate
the mean values of measured spectrum of these crops in their
growth stages and assess the data uncertainty.
In this study we classified the data extracted from the spectrum
database by variety and growth stages. And we analyzed the
measured data of typical crops and presented the statistics data
on canopy spectral reflectance and leaf spectral reflectance by
their mean values and the uncertainties. We did statistical
analysis, obtaining statistical parameters such as mean values,
minimum values, maximum values and variance. In this way,
we calculated the mean values and variances for constructing
the initial values and uncertainties of model parameters. Thus
the calculated results are shown in Fig. 1 and Fig.2. The
important a priori knowledge could be used in calculating
spectral VI and LAI retrieval.
3.2 “Beijing-l”data process
With Maximum Likelihood Method, we classified the “Beijing-
1” images into six targets: water, forest, grass, crop, city and the
unused. We classified the three images (16-March-2006, 14-
May-2006 and 17-June-2006) respectively. Comparing these
three classification images with Beijing LULC data, we used the
14-May-2006 classification image as the reference to make sure
the certain ground species. With this method, we obtain our
final classification image (Fig.3). As is shown in Fig.3, there are
7 classification classes, of which the winter wheat and summer
maize are reclassified from crop.
We got the needed a priori spectral knowledge of winter wheat
and summer maize in three spectral band (Green: 523nm —
605nm, Red: 630nm — 690nm and Near Infrared: 774nm —
900nm) and achieved the transformation from narrow band to
broad band in order to get the corresponding vegetation index,
such as RVI,NDVI,SAVI and so on. The arithmetic for
transformation is seen in Equation (1)
b lb
/X“' 1 (1)
A=a / A=a
where A is wavelength, [a, b] stands for the wavelength
bound, a a is the band response function for “Beijing-1”
image, py is the ground high spectrum reflectance on the
wavelength of A.
With this method, we obtained the statistical relationship
between ground based vegetation indexes and LAI, and
discovered better parameter inversion way. Then we simply
applied this ground relationship to the “Beijing-1” images.
4. RESULTS AND DISCUSSION
4.1 The statistical results of the measured spectral
reflectance in time sequence
Fig.l and Fig.2 include all crop species from measured crop
spectral reflectance data based on the Spectral Database System
of Typical Objects in China. In two figures, each colored lines
show the mean values of leaf and canopy spectrum in different
crop growth stages. Along each colored line, there is a range in
black which stands for the variance calculated from the whole
measured data. The wider as the black depicted in the figure, the
larger the variances is.
Form Fig.l and Fig.2, we can see that the mean values of
measured canopy and leaf reflectance change in different
growth stages and the uncertainty in the near infrared region is
higher than in the visible region.
From Fig.l, we could see that each of rice, rape , maize and
winter wheat only have one-year data in 2002, 2004, 2003 and
2004, respectively, while cotton have two-year data in 2003 and
2004. The measured growth stages in 2003 are more than in
2004 for cotton. The leaf reflectance in green light and the near-
infrared region reach the peak in grain-filling stage for early rice,
while for the late rice, it is the ripen stage and milky stage,
respectively. The leaf reflectance in green and the near-infrared
bands reaches the peak in ripen stage and returning green stage
respectively for winter wheat. In general, the temporal changing
information of the five crops’ leaf reflectance agrees with
others’ work.
And from Fig.2, we could see that each of rice, rape and maize
only have one-year data in 2003, 2004 and 2003, respectively,
while cotton and winter wheat have two-year data in 2003, 2004
and 2001, 2004, respectively. The measured growth stages in
2003 are more than in 2004 for cotton, and for winter wheat the
measured growth stages in 2004 are more than in 2001.The
canopy reflectance in green light reach the peak in flourishing
flowering stage for rape, in flourishing flowering stage and
flourishing belling stage for cotton, in early stage for maize and
winter wheat under the effect of soil. In general, the temporal